Tutorial for Lecture 3 - Analysis of COVID-19 Data
Scientific Data Analysis 6003ESC
Dr. Ido Bar
08/08/2021
Instructions
The following tutorial will let you reproduce the plots that we are going to create at the workshop using R.
Please read carefully and follow the steps. Wherever you see the Code icon on the right you can click on it to see the actual code used in that section.
Introduction
This tutorial will focus on analysing the updated data of the worldwide Novel Corona virus (COVID-19) pandemic.
There are several data sources available online. We will use the data collected from a range of official sources and hosted on the Our World in Data website (Mathieu et al. 2021).
Analyse Data in R
To run R and RStudio on Binder, click on this badge - 
.
Start RStudio and create a new project named Workshop3 in a new folder (if you need a reminder ho to do it, check out Workshop1 Tutorial on BB).
Once RStudio restarts inside the project’s folder, create a new R script named Workshop3.R and 2 new folders, one named data for our input data and another named output for our plots.
Install Extra Packages
For this analysis we will again use some packages from the Tidyverse, but this time we load the specific packages (which are supposed to be pre-installed on your computers) to try and avoid having to download the entire tidyverse. In addition to the Tidyverse packages we’ve got to know in the previous workshop we will use the plotly package to create interactive plots, paletteer for custom color palettes, readxl to read MS-Excel file, scales to format large numbers, lubridate to better handle dates, glue to paste together strings, patchwork to include several plots in a single figure and a few others to assist in getting the data into shape.
To install these packages, we will introduce a package called pacman that will assist in loading the required packages and installing them if they’re not already installed. To install it we use the install.packages('pacman') command, please note that the package name need to be quoted and that we only need to perform it once, or when we want or need to update the package. Once the package was installed, we can load its functions using the library(pacman) command and then load/install all the other packages at once with p_load() function.
# install required packages - needed only once! (comment with a # after first use)
install.packages('pacman')
# load required packages
library(pacman)
p_load(dplyr, tidyr, ggplot2, readr, paletteer, glue, scales, plotly, lubridate, patchwork, visdat)More information on installing and using R packages can be found in this tutorial.
Read Data
Now that we’ve got RStudio up and running and our packages installed and loaded, we can read data into R from our local computer or from web locations using dedicated functions specific to the file type (.csv, .txt, .xlsx, etc.).
We will use the read_csv() command/function from the readr package (part of the tidyverse) to load the data directly from a file on the the Our World in Data website into a variable of type data frame (table). If we don’t want to use external packages, we can use the read.csv() function from base R, which won’t automatically parse columns containing dates and in previous versions of R (< 4.0) will slightly change the structure of the resulting data frame (all text columns will be converted into factors).
> Note that in this case, we need to specify the column types because the data contains a lot of missing values that interfere with the automatic parsing of the column types._
Data Exploration
Let’s use built-in functions for a brief data exploration, such as head() to show the first 10 rows of the data and str() for the type of data in each column:
## # A tibble: 6 x 60
## iso_code continent location date total_cases new_cases new_cases_smoot~
## <chr> <fct> <chr> <date> <dbl> <dbl> <dbl>
## 1 AFG Asia Afghanistan 2020-02-24 1 1 NA
## 2 AFG Asia Afghanistan 2020-02-25 1 0 NA
## 3 AFG Asia Afghanistan 2020-02-26 1 0 NA
## 4 AFG Asia Afghanistan 2020-02-27 1 0 NA
## 5 AFG Asia Afghanistan 2020-02-28 1 0 NA
## 6 AFG Asia Afghanistan 2020-02-29 1 0 0.143
## # ... with 53 more variables: total_deaths <dbl>, new_deaths <dbl>,
## # new_deaths_smoothed <dbl>, total_cases_per_million <dbl>,
## # new_cases_per_million <dbl>, new_cases_smoothed_per_million <dbl>,
## # total_deaths_per_million <dbl>, new_deaths_per_million <dbl>,
## # new_deaths_smoothed_per_million <dbl>, reproduction_rate <dbl>,
## # icu_patients <dbl>, icu_patients_per_million <dbl>, hosp_patients <dbl>,
## # hosp_patients_per_million <dbl>, weekly_icu_admissions <dbl>, ...## spec_tbl_df [106,356 x 60] (S3: spec_tbl_df/tbl_df/tbl/data.frame)
## $ iso_code : chr [1:106356] "AFG" "AFG" "AFG" "AFG" ...
## $ continent : Factor w/ 6 levels "Asia","Europe",..: 1 1 1 1 1 1 1 1 1 1 ...
## $ location : chr [1:106356] "Afghanistan" "Afghanistan" "Afghanistan" "Afghanistan" ...
## $ date : Date[1:106356], format: "2020-02-24" "2020-02-25" ...
## $ total_cases : num [1:106356] 1 1 1 1 1 1 1 1 2 4 ...
## $ new_cases : num [1:106356] 1 0 0 0 0 0 0 0 1 2 ...
## $ new_cases_smoothed : num [1:106356] NA NA NA NA NA 0.143 0.143 0 0.143 0.429 ...
## $ total_deaths : num [1:106356] NA NA NA NA NA NA NA NA NA NA ...
## $ new_deaths : num [1:106356] NA NA NA NA NA NA NA NA NA NA ...
## $ new_deaths_smoothed : num [1:106356] NA NA NA NA NA 0 0 0 0 0 ...
## $ total_cases_per_million : num [1:106356] 0.026 0.026 0.026 0.026 0.026 0.026 0.026 0.026 0.051 0.103 ...
## $ new_cases_per_million : num [1:106356] 0.026 0 0 0 0 0 0 0 0.026 0.051 ...
## $ new_cases_smoothed_per_million : num [1:106356] NA NA NA NA NA 0.004 0.004 0 0.004 0.011 ...
## $ total_deaths_per_million : num [1:106356] NA NA NA NA NA NA NA NA NA NA ...
## $ new_deaths_per_million : num [1:106356] NA NA NA NA NA NA NA NA NA NA ...
## $ new_deaths_smoothed_per_million : num [1:106356] NA NA NA NA NA 0 0 0 0 0 ...
## $ reproduction_rate : num [1:106356] NA NA NA NA NA NA NA NA NA NA ...
## $ icu_patients : num [1:106356] NA NA NA NA NA NA NA NA NA NA ...
## $ icu_patients_per_million : num [1:106356] NA NA NA NA NA NA NA NA NA NA ...
## $ hosp_patients : num [1:106356] NA NA NA NA NA NA NA NA NA NA ...
## $ hosp_patients_per_million : num [1:106356] NA NA NA NA NA NA NA NA NA NA ...
## $ weekly_icu_admissions : num [1:106356] NA NA NA NA NA NA NA NA NA NA ...
## $ weekly_icu_admissions_per_million : num [1:106356] NA NA NA NA NA NA NA NA NA NA ...
## $ weekly_hosp_admissions : num [1:106356] NA NA NA NA NA NA NA NA NA NA ...
## $ weekly_hosp_admissions_per_million : num [1:106356] NA NA NA NA NA NA NA NA NA NA ...
## $ new_tests : num [1:106356] NA NA NA NA NA NA NA NA NA NA ...
## $ total_tests : num [1:106356] NA NA NA NA NA NA NA NA NA NA ...
## $ total_tests_per_thousand : num [1:106356] NA NA NA NA NA NA NA NA NA NA ...
## $ new_tests_per_thousand : num [1:106356] NA NA NA NA NA NA NA NA NA NA ...
## $ new_tests_smoothed : num [1:106356] NA NA NA NA NA NA NA NA NA NA ...
## $ new_tests_smoothed_per_thousand : num [1:106356] NA NA NA NA NA NA NA NA NA NA ...
## $ positive_rate : num [1:106356] NA NA NA NA NA NA NA NA NA NA ...
## $ tests_per_case : num [1:106356] NA NA NA NA NA NA NA NA NA NA ...
## $ tests_units : chr [1:106356] NA NA NA NA ...
## $ total_vaccinations : num [1:106356] NA NA NA NA NA NA NA NA NA NA ...
## $ people_vaccinated : num [1:106356] NA NA NA NA NA NA NA NA NA NA ...
## $ people_fully_vaccinated : num [1:106356] NA NA NA NA NA NA NA NA NA NA ...
## $ new_vaccinations : num [1:106356] NA NA NA NA NA NA NA NA NA NA ...
## $ new_vaccinations_smoothed : num [1:106356] NA NA NA NA NA NA NA NA NA NA ...
## $ total_vaccinations_per_hundred : num [1:106356] NA NA NA NA NA NA NA NA NA NA ...
## $ people_vaccinated_per_hundred : num [1:106356] NA NA NA NA NA NA NA NA NA NA ...
## $ people_fully_vaccinated_per_hundred : num [1:106356] NA NA NA NA NA NA NA NA NA NA ...
## $ new_vaccinations_smoothed_per_million: num [1:106356] NA NA NA NA NA NA NA NA NA NA ...
## $ stringency_index : num [1:106356] 8.33 8.33 8.33 8.33 8.33 ...
## $ population : num [1:106356] 38928341 38928341 38928341 38928341 38928341 ...
## $ population_density : num [1:106356] 54.4 54.4 54.4 54.4 54.4 ...
## $ median_age : num [1:106356] 18.6 18.6 18.6 18.6 18.6 18.6 18.6 18.6 18.6 18.6 ...
## $ aged_65_older : num [1:106356] 2.58 2.58 2.58 2.58 2.58 ...
## $ aged_70_older : num [1:106356] 1.34 1.34 1.34 1.34 1.34 ...
## $ gdp_per_capita : num [1:106356] 1804 1804 1804 1804 1804 ...
## $ extreme_poverty : num [1:106356] NA NA NA NA NA NA NA NA NA NA ...
## $ cardiovasc_death_rate : num [1:106356] 597 597 597 597 597 ...
## $ diabetes_prevalence : num [1:106356] 9.59 9.59 9.59 9.59 9.59 9.59 9.59 9.59 9.59 9.59 ...
## $ female_smokers : num [1:106356] NA NA NA NA NA NA NA NA NA NA ...
## $ male_smokers : num [1:106356] NA NA NA NA NA NA NA NA NA NA ...
## $ handwashing_facilities : num [1:106356] 37.7 37.7 37.7 37.7 37.7 ...
## $ hospital_beds_per_thousand : num [1:106356] 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 ...
## $ life_expectancy : num [1:106356] 64.8 64.8 64.8 64.8 64.8 ...
## $ human_development_index : num [1:106356] 0.511 0.511 0.511 0.511 0.511 0.511 0.511 0.511 0.511 0.511 ...
## $ excess_mortality : num [1:106356] NA NA NA NA NA NA NA NA NA NA ...
## - attr(*, "spec")=
## .. cols(
## .. iso_code = col_character(),
## .. continent = col_factor(levels = NULL, ordered = FALSE, include_na = FALSE),
## .. location = col_character(),
## .. date = col_date(format = ""),
## .. total_cases = col_double(),
## .. new_cases = col_double(),
## .. new_cases_smoothed = col_double(),
## .. total_deaths = col_double(),
## .. new_deaths = col_double(),
## .. new_deaths_smoothed = col_double(),
## .. total_cases_per_million = col_double(),
## .. new_cases_per_million = col_double(),
## .. new_cases_smoothed_per_million = col_double(),
## .. total_deaths_per_million = col_double(),
## .. new_deaths_per_million = col_double(),
## .. new_deaths_smoothed_per_million = col_double(),
## .. reproduction_rate = col_double(),
## .. icu_patients = col_double(),
## .. icu_patients_per_million = col_double(),
## .. hosp_patients = col_double(),
## .. hosp_patients_per_million = col_double(),
## .. weekly_icu_admissions = col_double(),
## .. weekly_icu_admissions_per_million = col_double(),
## .. weekly_hosp_admissions = col_double(),
## .. weekly_hosp_admissions_per_million = col_double(),
## .. new_tests = col_double(),
## .. total_tests = col_double(),
## .. total_tests_per_thousand = col_double(),
## .. new_tests_per_thousand = col_double(),
## .. new_tests_smoothed = col_double(),
## .. new_tests_smoothed_per_thousand = col_double(),
## .. positive_rate = col_double(),
## .. tests_per_case = col_double(),
## .. tests_units = col_character(),
## .. total_vaccinations = col_double(),
## .. people_vaccinated = col_double(),
## .. people_fully_vaccinated = col_double(),
## .. new_vaccinations = col_double(),
## .. new_vaccinations_smoothed = col_double(),
## .. total_vaccinations_per_hundred = col_double(),
## .. people_vaccinated_per_hundred = col_double(),
## .. people_fully_vaccinated_per_hundred = col_double(),
## .. new_vaccinations_smoothed_per_million = col_double(),
## .. stringency_index = col_double(),
## .. population = col_double(),
## .. population_density = col_double(),
## .. median_age = col_double(),
## .. aged_65_older = col_double(),
## .. aged_70_older = col_double(),
## .. gdp_per_capita = col_double(),
## .. extreme_poverty = col_double(),
## .. cardiovasc_death_rate = col_double(),
## .. diabetes_prevalence = col_double(),
## .. female_smokers = col_double(),
## .. male_smokers = col_double(),
## .. handwashing_facilities = col_double(),
## .. hospital_beds_per_thousand = col_double(),
## .. life_expectancy = col_double(),
## .. human_development_index = col_double(),
## .. excess_mortality = col_double()
## .. )Descriptive Statistics
We can also produce some descriptive statistics to better understand the data and the nature of each variable. The summary() function (as can be guessed by its name) provides a quick summary of basic descriptive statistics, such as the mean, min, max and quantiles for continuous numerical values.
## iso_code continent location
## Length:106356 Asia :25041 Length:106356
## Class :character Europe :25137 Class :character
## Mode :character Africa :27249 Mode :character
## North America:13320
## South America: 6344
## Oceania : 4326
## NA's : 4939
## date total_cases new_cases new_cases_smoothed
## Min. :2020-01-01 Min. : 1 Min. :-74347 Min. : -6223.0
## 1st Qu.:2020-07-14 1st Qu.: 1479 1st Qu.: 2 1st Qu.: 7.9
## Median :2020-11-27 Median : 15408 Median : 77 Median : 96.1
## Mean :2020-11-21 Mean : 1178243 Mean : 6146 Mean : 6149.9
## 3rd Qu.:2021-04-03 3rd Qu.: 163984 3rd Qu.: 845 3rd Qu.: 893.1
## Max. :2021-07-31 Max. :197872222 Max. :905993 Max. :826368.1
## NA's :4345 NA's :4348 NA's :5358
## total_deaths new_deaths new_deaths_smoothed
## Min. : 1 Min. :-1918.0 Min. : -232.143
## 1st Qu.: 59 1st Qu.: 0.0 1st Qu.: 0.000
## Median : 452 Median : 2.0 Median : 1.429
## Mean : 31203 Mean : 145.6 Mean : 131.506
## 3rd Qu.: 4340 3rd Qu.: 18.0 3rd Qu.: 14.714
## Max. :4217383 Max. :18056.0 Max. :14735.286
## NA's :14687 NA's :14532 NA's :5358
## total_cases_per_million new_cases_per_million new_cases_smoothed_per_million
## Min. : 0.0 Min. :-3162.163 Min. :-276.825
## 1st Qu.: 284.2 1st Qu.: 0.239 1st Qu.: 1.376
## Median : 2085.8 Median : 9.053 Median : 11.957
## Mean : 14563.7 Mean : 77.375 Mean : 77.486
## 3rd Qu.: 15638.2 3rd Qu.: 72.196 3rd Qu.: 80.325
## Max. :189969.6 Max. :18293.675 Max. :4083.500
## NA's :4886 NA's :4889 NA's :5894
## total_deaths_per_million new_deaths_per_million
## Min. : 0.001 Min. :-76.445
## 1st Qu.: 8.835 1st Qu.: 0.000
## Median : 58.104 Median : 0.137
## Mean : 316.492 Mean : 1.539
## 3rd Qu.: 354.097 3rd Qu.: 1.304
## Max. :5955.172 Max. :218.329
## NA's :15215 NA's :15060
## new_deaths_smoothed_per_million reproduction_rate icu_patients
## Min. :-10.921 Min. :-0.010 Min. : 0
## 1st Qu.: 0.000 1st Qu.: 0.840 1st Qu.: 29
## Median : 0.165 Median : 1.010 Median : 146
## Mean : 1.390 Mean : 1.009 Mean : 989
## 3rd Qu.: 1.274 3rd Qu.: 1.180 3rd Qu.: 652
## Max. : 63.140 Max. : 5.800 Max. :28889
## NA's :5894 NA's :20614 NA's :95339
## icu_patients_per_million hosp_patients hosp_patients_per_million
## Min. : 0.00 Min. : 0 Min. : 0.00
## 1st Qu.: 3.85 1st Qu.: 107 1st Qu.: 21.06
## Median : 14.06 Median : 588 Median : 73.14
## Mean : 24.42 Mean : 4406 Mean : 160.33
## 3rd Qu.: 38.34 3rd Qu.: 2544 3rd Qu.: 227.30
## Max. :192.74 Max. :133214 Max. :1532.57
## NA's :95339 NA's :93051 NA's :93051
## weekly_icu_admissions weekly_icu_admissions_per_million weekly_hosp_admissions
## Min. : 0.00 Min. : 0.00 Min. : 0.00
## 1st Qu.: 7.98 1st Qu.: 1.76 1st Qu.: 35.01
## Median : 39.93 Median : 7.65 Median : 249.68
## Mean : 249.43 Mean : 19.20 Mean : 3017.57
## 3rd Qu.: 190.31 3rd Qu.: 21.19 3rd Qu.: 1307.00
## Max. :4002.46 Max. :279.41 Max. :116323.00
## NA's :105377 NA's :105377 NA's :104475
## weekly_hosp_admissions_per_million new_tests total_tests
## Min. : 0.00 Min. :-239172 Min. : 0
## 1st Qu.: 7.56 1st Qu.: 1743 1st Qu.: 179821
## Median : 31.09 Median : 6450 Median : 912353
## Mean : 93.99 Mean : 49639 Mean : 8628308
## 3rd Qu.: 105.34 3rd Qu.: 24822 3rd Qu.: 3803333
## Max. :2656.91 Max. :3740296 Max. :486263680
## NA's :104475 NA's :59095 NA's :59369
## total_tests_per_thousand new_tests_per_thousand new_tests_smoothed
## Min. : 0.00 Min. : -6.32 Min. : 0
## 1st Qu.: 15.82 1st Qu.: 0.15 1st Qu.: 1840
## Median : 79.32 Median : 0.65 Median : 6932
## Mean : 352.03 Mean : 2.23 Mean : 47037
## 3rd Qu.: 329.17 3rd Qu.: 2.02 3rd Qu.: 27450
## Max. :11304.99 Max. :327.09 Max. :3080396
## NA's :59369 NA's :59095 NA's :51003
## new_tests_smoothed_per_thousand positive_rate tests_per_case
## Min. : 0.00 Min. :0.00 Min. : 1.1
## 1st Qu.: 0.16 1st Qu.:0.02 1st Qu.: 7.8
## Median : 0.69 Median :0.05 Median : 19.0
## Mean : 2.13 Mean :0.09 Mean : 163.2
## 3rd Qu.: 2.08 3rd Qu.:0.13 3rd Qu.: 58.1
## Max. :90.85 Max. :0.93 Max. :50000.0
## NA's :51003 NA's :54643 NA's :55275
## tests_units total_vaccinations people_vaccinated
## Length:106356 Min. :0.000e+00 Min. :0.000e+00
## Class :character 1st Qu.:1.556e+05 1st Qu.:1.272e+05
## Mode :character Median :1.187e+06 Median :8.280e+05
## Mean :4.255e+07 Mean :2.310e+07
## 3rd Qu.:7.564e+06 3rd Qu.:4.791e+06
## Max. :4.136e+09 Max. :2.206e+09
## NA's :86304 NA's :87184
## people_fully_vaccinated new_vaccinations new_vaccinations_smoothed
## Min. :1.000e+00 Min. : 0 Min. : 0
## 1st Qu.:6.086e+04 1st Qu.: 5033 1st Qu.: 875
## Median :5.033e+05 Median : 29148 Median : 7232
## Mean :1.308e+07 Mean : 745869 Mean : 356731
## 3rd Qu.:3.109e+06 3rd Qu.: 162009 3rd Qu.: 48229
## Max. :1.135e+09 Max. :48884282 Max. :43389920
## NA's :90112 NA's :89662 NA's :71279
## total_vaccinations_per_hundred people_vaccinated_per_hundred
## Min. : 0.00 Min. : 0.00
## 1st Qu.: 3.05 1st Qu.: 2.47
## Median : 15.35 Median : 10.91
## Mean : 30.43 Mean : 19.24
## 3rd Qu.: 46.65 3rd Qu.: 30.89
## Max. :232.72 Max. :116.73
## NA's :86304 NA's :87184
## people_fully_vaccinated_per_hundred new_vaccinations_smoothed_per_million
## Min. : 0.00 Min. : 0
## 1st Qu.: 1.35 1st Qu.: 417
## Median : 5.92 Median : 1895
## Mean : 13.04 Mean : 3398
## 3rd Qu.: 18.55 3rd Qu.: 4965
## Max. :115.99 Max. :118759
## NA's :90112 NA's :71279
## stringency_index population population_density median_age
## Min. : 0.00 Min. :4.700e+01 Min. : 0.137 Min. :15.10
## 1st Qu.: 43.52 1st Qu.:2.142e+06 1st Qu.: 36.253 1st Qu.:22.20
## Median : 59.72 Median :9.660e+06 Median : 83.479 Median :29.70
## Mean : 57.89 Mean :1.236e+08 Mean : 390.400 Mean :30.55
## 3rd Qu.: 74.07 3rd Qu.:3.347e+07 3rd Qu.: 209.588 3rd Qu.:39.10
## Max. :100.00 Max. :7.795e+09 Max. :20546.766 Max. :48.20
## NA's :17964 NA's :706 NA's :7657 NA's :11632
## aged_65_older aged_70_older gdp_per_capita extreme_poverty
## Min. : 1.144 Min. : 0.526 Min. : 661.2 Min. : 0.10
## 1st Qu.: 3.466 1st Qu.: 2.063 1st Qu.: 4466.5 1st Qu.: 0.60
## Median : 6.378 Median : 3.871 Median : 12951.8 Median : 2.20
## Mean : 8.781 Mean : 5.558 Mean : 19280.6 Mean :13.43
## 3rd Qu.:14.312 3rd Qu.: 8.678 3rd Qu.: 27216.4 3rd Qu.:21.20
## Max. :27.049 Max. :18.493 Max. :116935.6 Max. :77.60
## NA's :12692 NA's :12154 NA's :11227 NA's :42301
## cardiovasc_death_rate diabetes_prevalence female_smokers male_smokers
## Min. : 79.37 Min. : 0.990 Min. : 0.10 Min. : 7.70
## 1st Qu.:168.71 1st Qu.: 5.310 1st Qu.: 1.90 1st Qu.:21.60
## Median :242.65 Median : 7.110 Median : 6.30 Median :31.40
## Mean :258.74 Mean : 7.955 Mean :10.58 Mean :32.71
## 3rd Qu.:329.94 3rd Qu.:10.080 3rd Qu.:19.30 3rd Qu.:41.10
## Max. :724.42 Max. :30.530 Max. :44.00 Max. :78.10
## NA's :11286 NA's :8697 NA's :32041 NA's :33127
## handwashing_facilities hospital_beds_per_thousand life_expectancy
## Min. : 1.19 Min. : 0.100 Min. :53.28
## 1st Qu.: 19.35 1st Qu.: 1.300 1st Qu.:67.92
## Median : 49.84 Median : 2.400 Median :74.62
## Mean : 50.80 Mean : 3.027 Mean :73.24
## 3rd Qu.: 83.24 3rd Qu.: 3.861 3rd Qu.:78.74
## Max. :100.00 Max. :13.800 Max. :86.75
## NA's :58608 NA's :19813 NA's :5417
## human_development_index excess_mortality
## Min. :0.394 Min. :-95.59
## 1st Qu.:0.602 1st Qu.: 0.43
## Median :0.744 Median : 7.44
## Mean :0.727 Mean : 18.13
## 3rd Qu.:0.848 3rd Qu.: 23.94
## Max. :0.957 Max. :410.12
## NA's :11174 NA's :102683## # A tibble: 106,356 x 60
## iso_code continent location date total_cases new_cases new_cases_smoot~
## <chr> <fct> <chr> <date> <dbl> <dbl> <dbl>
## 1 OWID_WRL <NA> World 2021-07-31 197872222 506087 589881.
## 2 OWID_WRL <NA> World 2021-07-30 197366135 739604 585280.
## 3 OWID_WRL <NA> World 2021-07-29 196626531 658221 578041.
## 4 OWID_WRL <NA> World 2021-07-28 195968310 634006 564770.
## 5 OWID_WRL <NA> World 2021-07-27 195334304 613517 554161.
## 6 OWID_WRL <NA> World 2021-07-26 194720787 542415 542521.
## 7 OWID_WRL <NA> World 2021-07-25 194178372 435316 536420.
## 8 OWID_WRL <NA> World 2021-07-24 193743056 473879 534600.
## 9 OWID_WRL <NA> World 2021-07-23 193269177 688931 534579.
## 10 OWID_WRL <NA> World 2021-07-22 192580246 565323 521709.
## # ... with 106,346 more rows, and 53 more variables: total_deaths <dbl>,
## # new_deaths <dbl>, new_deaths_smoothed <dbl>, total_cases_per_million <dbl>,
## # new_cases_per_million <dbl>, new_cases_smoothed_per_million <dbl>,
## # total_deaths_per_million <dbl>, new_deaths_per_million <dbl>,
## # new_deaths_smoothed_per_million <dbl>, reproduction_rate <dbl>,
## # icu_patients <dbl>, icu_patients_per_million <dbl>, hosp_patients <dbl>,
## # hosp_patients_per_million <dbl>, weekly_icu_admissions <dbl>, ...What are the metadata columns that describe our observations?
continent
location
dateWhy do we have observations with the continent as NA?
# check which location have continent as NA
covid_data %>% filter(is.na(continent)) %>% count(location)## # A tibble: 9 x 2
## location n
## <chr> <int>
## 1 Africa 535
## 2 Asia 557
## 3 Europe 556
## 4 European Union 556
## 5 International 541
## 6 North America 557
## 7 Oceania 554
## 8 South America 526
## 9 World 557Some rows contain summarised data of entire continents/World, we'll need to remove thoseWe can see that most of our data contains ‘0’ (check the difference between the median and the mean in total_cases and total_deaths columns). Just to confirm that, let’s plot a histogram of all the confirmed cases
ggplot(covid_data, aes(x=total_cases)) +
geom_histogram(fill="lightskyblue") +
theme_bw(def_text_size)
The data is evolving over days (a time-series), to there’s no point treating it as a random population sample.
Time-series plot
Let’s look at confirmed cases and total deaths data for the 10 most affected countries (to date). To find out these countries so we need to wrangle our data a little bit using the following steps:
- First we remove all observations for combined continents with
filter(!is.na(continent) - Then we group it by location with
group_by() - Then we sort it within each location by date (from latest to earliest) with
arrange(desc(date)) - We select just the most recent data point from each location with
slice(1)and remove grouping withungroup() - Next we arrange it by descending order of total deaths and select the top 10 observations (one for each location)
- Finally, we subset our original data to contain just the countries from our vector with
inner_join()
Optional step:
- We can recode the
locationvariable as a factor and order it so the countries will be ordered in the legend by the number of cases with
Then we can look at the data as a table and make a plot with the number of cases in the y-axis and date in the x-axis.
# find the 10 most affected countries (to date)
latest_data <- covid_data %>% filter(!is.na(continent)) %>%
group_by(location) %>% arrange(desc(date)) %>% slice(1) %>% ungroup()
most_affected_countries <- latest_data %>%
arrange(desc(total_deaths)) %>% slice(1:10) %>%
select(location)
# subset just the data from the 10 most affected countries and order them from the most affected to the least one
most_affected_data <- covid_data %>%
inner_join(most_affected_countries) %>%
mutate(Country=factor(location, levels = most_affected_countries$location))
# create a line plot the data of total cases
ggplot(most_affected_data, aes(x=date, y=total_cases, colour=Country)) +
geom_line(size=0.75) + scale_y_continuous(labels=comma) +
scale_color_paletteer_d("rcartocolor::Bold") +
labs(color="Country", y = "Total COVID-19 cases") +
theme_bw(def_text_size)
It’s a bit hard to figure out how the pandemic evolved because the numbers in US, Brazil and India are an order of magnitude larger than the rest (which are very close to each other). How can we make it more visible (and also improve how of the dates appear in the x-axis)?
# better formatting of date axis, log scale
plot <- ggplot(most_affected_data, aes(x=date, y=total_cases, colour=Country)) +
geom_line(size=0.75) + scale_y_log10(labels=comma) +
scale_x_date(NULL,
breaks = breaks_width("2 months"),
labels = label_date_short()) +
scale_color_paletteer_d("rcartocolor::Bold") +
labs(color="Country", y = "Total COVID-19 cases") +
theme_bw(def_text_size)
# show an interactive plot
ggplotly(plot)Why did we get a warning message and why the graphs don’t start at the bottom of the x-axis? How can we solve it? What can we infer from the graph (exponential increase)?
What happens when we take the log of 0?? Can we remove those 0s with the `filter()` function (or add a very small number to them)?
We can see a very similar trend for most countries and while the curve has flattened substantially in April last year, the numbers are still rising. It is also evident that Europe got hit by a second wave arount October last year and India in April this year.Total deaths
Let’s have a look at the total deaths in these countries (and get rid of the minor grid lines to make Frank happy)
# create a line plot the data of total deaths
ggplot(most_affected_data, aes(x=date, y=total_deaths, colour=Country)) +
geom_line(size=0.75) + scale_y_continuous(labels=comma) +
scale_x_date(NULL,
breaks = breaks_width("2 months"),
labels = label_date_short()) +
scale_color_paletteer_d("rcartocolor::Bold") +
labs(color="Country", y = "Total deaths") +
theme_bw(def_text_size) +
theme(panel.grid.minor = element_blank()) # remove minor grid lines
Vaccination rates
Let’s have a look at the number of vaccinated people.
# vaccination rates
ggplot(most_affected_data, aes(x=date, y=people_vaccinated, colour=Country)) +
geom_line(size=0.75) + scale_y_continuous(labels=comma) +
scale_color_paletteer_d("rcartocolor::Bold") +
scale_x_date(NULL,
breaks = breaks_width("2 months"),
labels = label_date_short()) +
labs(color="Country") +
theme_bw(def_text_size) +
theme(panel.grid.minor = element_blank())
The graphs are “broken”, meaning that it is not continuous and we have some missing data.
Let’s visualise some of the variables in our data and assess “missingness”.
# visualise missingness
vis_dat(covid_data %>% filter(date>dmy("01-01-2021")) %>%
select(continent, location, total_cases, total_deaths,
hosp_patients, people_vaccinated, people_fully_vaccinated))
# find which countries has the most number of observations (least missing data)
covid_data %>% filter(!is.na(continent), !is.na(people_vaccinated)) %>% # group_by(location) %>%
count(location) %>% arrange(desc(n)) %>% print(n=30)## # A tibble: 215 x 2
## location n
## <chr> <int>
## 1 Norway 240
## 2 Canada 230
## 3 Israel 225
## 4 Denmark 224
## 5 Latvia 221
## 6 Liechtenstein 220
## 7 Switzerland 220
## 8 Czechia 218
## 9 Austria 217
## 10 Chile 217
## 11 Estonia 217
## 12 Italy 217
## 13 Lithuania 217
## 14 Slovenia 217
## 15 Bahrain 216
## 16 Germany 216
## 17 France 215
## 18 Belgium 214
## 19 Romania 210
## 20 Slovakia 208
## 21 United States 207
## 22 United Kingdom 206
## 23 Portugal 202
## 24 Greece 197
## 25 Bulgaria 196
## 26 Malta 194
## 27 Mexico 194
## 28 India 190
## 29 Argentina 189
## 30 Poland 182
## # ... with 185 more rowscovid_data %>% filter(!is.na(continent), !is.na(hosp_patients)) %>% # group_by(location) %>%
count(location) %>% arrange(desc(n)) %>% print(n=30)## # A tibble: 29 x 2
## location n
## <chr> <int>
## 1 France 525
## 2 Estonia 519
## 3 Italy 518
## 4 Israel 516
## 5 Portugal 516
## 6 Sweden 515
## 7 Netherlands 513
## 8 Czechia 512
## 9 Canada 509
## 10 Hungary 508
## 11 Cyprus 504
## 12 Ireland 503
## 13 Slovenia 503
## 14 Belgium 498
## 15 Luxembourg 497
## 16 United Kingdom 490
## 17 Latvia 482
## 18 Austria 481
## 19 Bulgaria 476
## 20 Poland 460
## 21 Croatia 453
## 22 Slovakia 452
## 23 Norway 450
## 24 Denmark 415
## 25 United States 375
## 26 Iceland 351
## 27 Finland 315
## 28 Spain 233
## 29 Lithuania 216Hospitalisation and Vaccination rates
Now we can focus on a subset of countries that have more complete vaccination and hospitalisation rates data, so we could compare them.
countries_subset <- c("Italy", "United States", "Israel", "United Kingdom", "France", "Canada", "Czechia")
# subset our original data to these countries
hosp_data <- covid_data %>% filter(location %in% countries_subset)
# define a new colour palette for these countries
col_pal <- "ggsci::category10_d3"Let’s look at hospitalisation rates first.
ggplot(hosp_data, aes(x=date, y = hosp_patients,colour=location)) +
geom_line(size=0.75) + scale_y_continuous(labels=comma) +
scale_color_paletteer_d(col_pal) +
scale_x_date(name = NULL,
breaks = breaks_width("2 months"),
labels = label_date_short()) +
labs(color="Country", y = "Hospitalised patients") +
theme_bw(def_text_size) +
theme(panel.grid.minor = element_blank())
Can you identify the “waves” in each country?
It’s hard to see the details in the countries with lower number of hospitalised patients, how can we improve the visualisation?
Look at hospitalision rates proportional to the population size!# hosp per population size
p1 <- ggplot(hosp_data,
aes(x=date, y = hosp_patients_per_million,colour=location)) +
geom_line(size=0.75) +
scale_y_continuous(labels=comma) +
scale_color_paletteer_d(col_pal) +
scale_x_date(name = NULL,
breaks = breaks_width("2 months"),
labels = label_date_short()) +
labs(color="Country", y = "Hospitalised patients (per million)") +
theme_bw(def_text_size) +
theme(panel.grid.minor = element_blank())
p1
Now let’s try to compare this to vaccination rates
# total vaccination per population
p2 <- ggplot(hosp_data,
aes(x=date, y = people_fully_vaccinated_per_hundred ,colour=location)) +
geom_line(size=0.75, linetype="dashed") +
guides(color = guide_legend(override.aes = list(linetype="solid") ) ) +
scale_y_continuous(labels=comma) +
scale_color_paletteer_d(col_pal) +
scale_x_date(name = NULL,
breaks = breaks_width("2 months"),
labels = label_date_short()) +
labs(color="Country", y = "Fully vaccinated (per hundred)") +
theme_bw(def_text_size) +
theme(panel.grid.minor = element_blank())
p2
What will be the best way to compare these values?
Maybe like this:
(p1 + guides(color=FALSE)) / (p2 + theme(legend.position = "bottom")) #+ plot_layout(guides = 'collect')# show graphs on top of each other
Any suggestions?
There's a lot of empty "real estate" in the vaccination graph, maybe we could trim off 2020?p3 <- ggplot(hosp_data,
aes(x=date, y = people_fully_vaccinated_per_hundred ,colour=location)) +
geom_line(size=0.75, linetype="dashed") +
guides(color = guide_legend(override.aes = list(linetype="solid") ) ) +
scale_y_continuous(labels=comma) +
scale_color_paletteer_d(col_pal) +
scale_x_date(name = NULL,
limits = c(dmy("01-01-2021"), NA),
breaks = breaks_width("2 months"),
labels = label_date_short()) +
labs(color="Country", y = "Fully vaccinated (per hundred)") +
theme_bw(def_text_size) +
theme(panel.grid.minor = element_blank())
(p1 + guides(color=FALSE)) + p3 + plot_layout(guides = 'collect', widths = c(2, 1)) # maybe like this?
Let’s try to present them on the same graph (note the trick with the secondary y-axis).
# show on the same graph
p4 <- ggplot(hosp_data,
aes(x=date, colour=location)) +
geom_line(aes(y = hosp_patients_per_million), size=0.75) +
geom_line(aes(y = people_fully_vaccinated_per_hundred*10), size=0.75, linetype="dashed") +
scale_y_continuous(labels=comma, name = "Hospitalised patients per million (solid)",
# Add a second axis and specify its features
sec.axis = sec_axis(trans=~./10, name="Fully vaccinated per hundred (dashed)")) +
scale_color_paletteer_d(col_pal) +
scale_x_date(NULL,
breaks = breaks_width("2 months"),
labels = label_date_short()) +
labs(color="Country") +
theme_bw(def_text_size) +
theme(panel.grid.minor = element_blank())
p4
Probably best to present them on the same graph (note the trick with the secondary y-axis), but for each country separately (done with the facet_wrap() function).
# show on the same graph, but separate each country
p4 +
scale_x_date(NULL,
breaks = breaks_width("4 months"),
labels = label_date_short()) +
facet_wrap(~location)
Anything to worry about???
Save the plot to a folder.
# create output folder
dir.create("./output", showWarnings = FALSE)
# save the plot to pdf file
ggsave("output/hospit_vacc_rates_facet_country.pdf", width=14, height = 8)Questions
- What other variables we could analyse?
- Any correlated variables?
- What we should take into account that might bias the results or the true status of the pandemic?
1. Mortalities (Case Fatality Rate)?
2. Suggestions? (cases per population density, vaccination rates by country income, deaths by number of beds per capita, etc.
3. Level of reporting in each country... Additional Resources
- Johns Hopkins University Center for Systems Science and Engineering (JHU CSSE) data repository and website
- EU 14-days COVID-19 data for download in CSV/Excel format link
- Be awesome in ggplot2: A Practical Guide to be Highly Effective – R software and data visualization (link)
- The R graph gallery - From the creator of “Data to Viz” - a comprehensive gallery of R charts, with reproducible code examples (link)
- COVID-19 vaccination data in Our World in Data site
- My very own COVID-19 dashboard (created in R, needs updating)
Contact
Please contact me at i.bar@griffith.edu.au for any questions or comments.
References
Mathieu E, Ritchie H, Ortiz-Ospina E, et al. (2021) A global database of COVID-19 vaccinations. Nat Hum Behav 5:947–953. doi: 10.1038/s41562-021-01122-8